Introduction

The electrochemical reduction of carbon dioxide is of great interest regarding the present day environmental challenge as it offers a potential way to recycle CO2 into valuable products for energy needs or for industrial applications. In particular, the production from CO2 of a gas mixture that could provide direct energy is a very attractive prospect. Therefore, many studies have been devoted to the finding of materials and conditions that could match the requirements for these purposes [1–5].

The electroreduction of CO2 in aqueous solutions has been investigated at various metals, usually taking place at high overpotentials and yielding gaseous carbon monoxide and aqueous formate as final products [6, 7]. Pure copper distinguishes itself by its efficient ability to produce a variety of gaseous and aqueous products such as hydrocarbons and alcohols. Since the pioneering work of Hori [8], copper has attracted a considerable attention, and much of the advancements in the study of the aqueous electrochemical reduction of CO2 at copper electrodes have been summarized up to 2006 in a review published by Gattrell et al. [9]. It is now well established that the product distribution depends greatly on the surface state [10–12] and crystallographic orientation of the polycrystalline copper electrode [13–16], though the underlying mechanisms are not well understood. Recently, the group of Koper [17] has followed, by means of online mass spectrometry, the formation and consumption of intermediates during the electroreduction of carbon dioxide on copper electrodes in a phosphate buffer (pH 7) and has proposed a new mechanism for the selectivity to C1 and C2 species.

In addition to the voltammetric and chromatographic methods that were frequently used to examine the activity of electrodes towards the electroreduction of CO2, in situ IR spectroscopic measurements were also performed in order to detect intermediates and adsorbed species formed at the electrode during the reduction process. In this respect, the group of Hori [18] reported the presence of adsorbed CO at the copper electrode whereas Hernandez et al. [19] could only evidence the conversion of CO2 to HCO −3 and CO 2−3 due to alkaline conditions near the electrode. The adsorption of CO on low-index copper single crystals has been recently revisited in the presence of a weakly adsorbing electrolyte [20]. From a combined in situ infrared spectroscopy, cyclic voltammetry and density functional theory (DFT) calculations study, the Schiffrin group concludes that the electrocatalytic behaviour of Cu electrodes for CO2 reduction needs to be re-evaluated to take into account the CO adsorption and its possible co-adsorption with the electrolyte anion.

Although copper promotes the formation of various products, the high overpotentials required for the electroreduction remain a problematic issue in terms of yield and energy efficiency. Trends in binding energies for the intermediates in CO2 electroreduction have become available from DFT calculations [21, 22], forming the basis of new strategies, presented recently by Peterson and Nørskov [23], that may lead to reduced overpotentials for the electroreduction of carbon dioxide. These are relying on the fact that effective catalysts must be able to efficiently catalyse the protonation of adsorbed CO or COH and have poor activity for the hydrogen evolution reaction. One proposed strategy in this respect consists in alloying Cu with metals having a higher oxygen affinity so that intermediates such as CHO may bind to the surface through both the carbon and oxygen atoms. The stability of CHO is then increased without affecting the stability of CO. In the past Watanabe et al. [24] have already investigated the catalytic properties of various copper-based bimetallic electrodes. In some cases the catalytic activity of the electrode is improved, such as Cu–Ni, Cu–Sn or Cu–Pb alloys, but in other cases the catalytic aptitudes of the metals are diluted, as it happens for Cu–Ag and Cu–Cd electrodes. So far, only one work has been devoted to the study of copper–gold bimetallic electrodes. Kyriacou and Anagnostopoulos [25] used Au-modified Cu electrodes and observed that the production of CH4 falls when the surface contains more gold. The proportion of Au at the surface of the different electrodes did not exceed 7.2 %.

In the present work, we report on the electroactivity of well-characterized copper-based surfaces for the CO2 electroreduction. Information was obtained at room temperature and in an aqueous phosphate buffer from cyclic voltammetry and long-term electrolyses under potentiostatic conditions, contrary to most earlier published works where electrolyses were performed at constant current (i.e. galvanostatic conditions). Moreover, we focus our investigation on the production of gaseous products, because only very small amounts of water soluble products were detected.

First we are presenting experimental data obtained with copper single-crystal electrodes allowing us to correlate the activity of the Cu electrodes to the atomic arrangement of the surface in the light of recent experimental and theoretical studies on intermediates and more specifically on CO adsorption. In the second part, we show new data relative to the use of bimetallic catalysts where copper is alloyed with gold, a metal for which the binding energy of CHO is strengthened relative to that of CO [23].

Experimental

Chemicals

The supporting electrolytes consisted of KH2PO4 (Merck, Pro Analysi)/K2HPO4 (Merck, Pro Analysi) buffer solutions prepared with ultra pure water from a Milli-Q system (Millipore). All gases (CO2 and N2 from Air Liquide and the GC standards CO, CH4, C2H4 and C2H6 from Alltech) were used as received.

Electrode Preparation

Copper electrodes of various crystallographic orientations and copper–gold alloys of diverse compositions were used as working electrodes. The pure Cu electrodes were Cu (poly), Cu (111) and Cu (100). The home-made single-crystal electrodes were grown by the vertical Bridgman method. The Laüe back-scattering X-ray diffraction method (Philips PW1729) was employed to characterize and orientate the single-crystal surfaces. The Cu electrodes used for the cyclic voltammetry experiments were electropolished in 70 % H3PO4 (1.8 V, 10 min). For all other measurements (electrolysis measurements), a mechanical polishing was applied: the electrodes were first polished with SiC abrasive paper (P4000, Struers) then with 1-μm alumina–water slurry on a smooth polishing cloth (Struers). The electrodes were sonicated and rinsed with Milli-Q water.

The gold-containing electrodes were Au1Cu99, Au10Cu90, Au20Cu80, Au50Cu50 and Au. In the Au x Cu100 − x denomination, x represents the bulk composition of the alloy, expressed in atomic percent of gold. Copper 99.999 % was purchased from Alfa Aesar and gold 99.999 % from Advent. Cu and Au were alloyed by melting both metals together in various proportions, according to a procedure described elsewhere [26]. We have reported earlier [26] on the segregation of gold at the Au x Cu100 − x surface at high temperatures (800–1,000 K) and the depletion of copper at the surface after electrooxidation in aqueous solutions. To avoid differences between the surface and bulk compositions, no thermal annealing or electrochemical polishing was performed, the only treatment of the alloys consisting in mechanical polishing. Prior to every measurement, the alloy and gold surfaces were first polished with SiC abrasive paper (P4000, Struers) then with 1-μm alumina–water slurry on a smooth polishing cloth (Struers). The electrodes were sonicated and rinsed with Milli-Q water. The surface composition of the alloys was examined by means of Auger electron spectroscopy and found to be identical to the bulk composition, as expected from the simple removal of surface atomic layers by mechanical polishing. To ensure that the experimental conditions do not affect the surface composition, the latter was determined before and after the electrochemical experiments, without noticing any significant change.

Electrochemical Measurements

All electrochemical measurements were performed in a three-electrode cell connected to an Autolab PGSTAT 30 (Metrohm-Autolab). The hanging meniscus technique allowed only the face of interest of the working electrodes to be in contact with the electrolyte. The large area Pt counter electrode was separated from the main compartment by a glass frit. A Ag|AgCl|KCl (saturated) reference electrode (Radiometer XR300) completed the circuit. All potentials are reported with respect to this reference electrode.

Electrolysis

The electrolysis experiments were conducted in a home-designed gas tight cell. The solution was saturated with CO2, introduced into the electrolyte just before the experiment. During the electrolysis, the flow of CO2 was stopped and the electrolyte was kept under constant stirring. Unless otherwise stated, the electrolysis potential, time and temperature were fixed at −1.9 V, 3 h and 20 °C, respectively. The i–t curves were disturbed by the permanent stirring of the solution and the hydrogen evolution reaction (forming bubbles at the surface). The average current density during the electrolysis was between 11 and 24 mA/cm2 for all experiments. These conditions led to the accumulation of sufficient amounts of products for their detection and quantification.

At the end of the electrolysis, the gaseous phase was immediately sampled through the septum of the cell and analysed by gas chromatography. The gas chromatograph was a 6890-N apparatus (Agilent) equipped with a Carboxen 1000 filled column (Supelco) and both thermal conductivity detector and flame ionization detector. Setup parameters were controlled by the GC Chemstation software. Ionic chromatography and gas chromatography techniques were used to detect products of the CO2 electroreduction in the aqueous phase. Only very small quantities (<1 % of the total amount of products) of formate (on copper, copper–gold alloys and gold substrates) and ethanol (on copper electrodes) have been detected. These aqueous products were not considered in this work.

Results and Discussion

For the electrochemical reduction of CO2, its acidic properties have to be carefully considered. The choice of pH of the electrolyte is the result of a compromise, since the pH has to be low enough to minimize the conversion of dissolved CO2 to hydrogen carbonate and carbonate and high enough to ensure a limited contribution from the hydrogen evolution reaction (HER) taking place at the electrode surface. As a result, the experiments have been carried out at pHs close to 7.

Electroactivity of Copper Surfaces

Figure 1 presents the cyclic voltammogram of Cu (poly) recorded between +0.2 and −1.1 V in the absence and presence of CO2 dissolved in phosphate buffer electrolyte solutions. The voltammogram is characterized by a pair of peaks situated at ca. −0.1 V corresponding to the copper oxidation and subsequent reduction process. At the negative potential limit, a large reduction current is observed in the presence of CO2, but a closer examination of the voltammograms reveals that the addition of CO2 into the electrolyte shifts all the j–E response in the positive direction, because of the pH variation: when CO2 is dissolved in the electrolyte, the pH of the solution decreases from 7.8 to 6.5. Therefore, for the sake of comparison, the reference scans in the absence of CO2 were recorded in a N2-saturated 0.1 M KH2PO4/0.02 M K2HPO4 solution at pH 6.5. It can be seen that at identical pH, the reduction of carbon dioxide overlaps with the HER.

Fig. 1
figure 1

Cyclic voltammograms of Cu (poly) in N2-saturated 0.01 M KH2PO4/0.1 M K2HPO4, pH 7.8 (dotted lines), CO2-saturated 0.01 M KH2PO4/0.1 M K2HPO4 (solid line) and N2-saturated 0.1 M KH2PO4/0.02 M K2HPO4, pH 6.5 (dashed lines). Scan rate 20 mV/s

Full size image

Cyclic voltammograms for polycrystalline Cu and single-crystal Cu (111) electrodes recorded in the absence and presence of CO2 are shown in Fig. 2. In order to avoid changes in surface oxidation as well as surface roughening of the copper, the excursion of the electrode potential for all other CV measurements was kept between −0.8 and −1.5 V.

Fig. 2
figure 2

Cyclic voltammograms of Cu electrodes in CO2-saturated 0.01 M KH2PO4/0.1 M K2HPO4 (solid lines) and N2-saturated 0.1 M KH2PO4/0.02 M K2HPO4 (dashed lines). a, b Freshly electropolished electrodes and c, d after several cycles to +0.2 V. Scan rate 20 mV/s

Full size image

On the freshly electropolished electrodes (Fig. 2a, b), the presence of CO2 induces a significant shift of the HER to more negative potentials for the polycrystal electrode, but not for Cu (111). Such a shift has been already observed in the past [27] and was attributed to the presence of adsorbed CO. This interpretation is strongly supported by the recent results of Shaw et al. [20], who compared the electrochemical behaviour of CO at Cu (111), Cu (100) and Cu (110) electrodes. The authors concluded from their experimental data and DFT calculations that CO is adsorbed at a high enough coverage to inhibit the HER on the (110) and (100) faces of copper, contrary to the (111) surface where the HER remains unaffected, in very good agreement with our observations. These results show that the reduction of CO2, involving adsorbed CO, is very sensitive to the surface state. A further confirmation of this high sensitivity was obtained by performing an additional experiment, in which the potential was swept to +0.2 V and scanned back to −0.8 V, then recording a new voltammogram in the reduction region of CO2. This potential excursion in the oxidation region of the copper electrode induces a roughening of the electrodes, resulting in significant changes in the voltammograms obtained at Cu (poly) (Fig. 2c) and Cu (111) (Fig. 2d) electrodes. On the latter, the HER is now shifted in the negative direction, confirming that roughening has occurred. On polycrystalline copper, a peak is clearly evidenced at potentials around −1.3 V. This peak was also noticeable, though less pronounced, on the freshly prepared electrode (Fig. 2a). Such a peak was also observed by Hori et al. [28] and attributed to an inhibition of the reduction due to adsorbed CO blocking the Cu surface, before being further reduced at more negative potentials. Thus, the voltammetric results indicate that the reduction of CO2 is promoted at smooth single-crystal surfaces as compared to other surfaces (polycrystal or rough surfaces).

In order to further investigate the influence of the atomic arrangement of the Cu surfaces on their activity for CO2 electroreduction, we performed potentiostatic electrolysis experiments and analysed the products in the gaseous phase. In addition to H2, four species were identified in the gaseous phase after electrolyses: CO, CH4, C2H4 and traces of C2H6. Under potentiostatic conditions, it is expected that the reduction products will depend on the applied potential. This is confirmed by electrolysis experiments conducted under the same conditions on polycrystalline Cu at −1.5, −1.9 and −2.3 V. The results of the chromatographic analyses are presented in Table 1. At low overvoltage, −1.5 V, essentially carbon monoxide is formed in the gas phase. At more negative potentials, the proportion of CO decreases markedly in favour of CH4 which becomes by far the major product at −2.3 V. On the basis of these results, the intermediate potential −1.9 V turns out to be the most appropriate to explore the possible influence of the crystallography and composition of the electrodes and was thus used in the subsequent experiments.

Table 1 Analysis of the carbon-containing gaseous products after the electrochemical reduction of CO2 on Cu (poly) in 0.01 M KH2PO4/0.1 M K2HPO4
Full size table

The amounts of carbon-containing gaseous products detected after CO2 electroreduction at Cu surfaces are plotted in Fig. 3. The main products are clearly CH4 and CO. In our experiments, a C2 pathway is also operative but to a much lesser extent. The mechanism for the formation of methane as the product of a C1 pathway has been discussed recently by the group of Koper [17]. The authors suggest CHO(ads) as intermediate, in contrast to older studies where surface carbene CH2(ads) was considered as the key intermediate. Analysis of the data indicates that the fraction of methane increases in the order Cu (poly) < Cu (100) < Cu (111), whereas a concomitant decrease of the carbon monoxide fraction is observed. The absolute total amount of gaseous products increases in the same order. Furthermore the selectivity for CH4 seems to be enhanced on high-density surfaces. These results are consistent with a mechanism involving reactions between adsorbed species, i.e. adsorbed CO and adsorbed H. According to such a mechanism, the formation of hydrogenated products requires the simultaneous presence of adsorbed hydrogen and adsorbed carbon monoxide within a short distance allowing them to react together. The reduction of CO2 is thus expected to be dependent on the surface crystallography: the probability of finding adsorbed CO and H on adjacent copper sites is higher on high-density surfaces, like the (111) face, than on rough surfaces of lower atomic density. Consequently the production of CH4 is enhanced at these smooth, high-density surfaces. Thermodynamic data from surface science studies give additional support to this interpretation. The binding energy (E b) of CO on copper single crystals measured by thermal desorption spectroscopy experiments has been reported by Vollmer et al. [29]. The adsorption of CO on Cu surfaces is favoured according to the following order: Cu (111) (E b = 47 kJ mol−1) < Cu (100) (E b = 51 kJ mol−1) < Cu (110) (E b = 54 kJ mol−1) < Cu (poly) (E b = 58 kJ mol−1). Such sequence of adsorption strength is compatible with the feature observed in the voltammetric curves. To explain the selectivity of low-index faces towards methane formation, the adsorption of hydrogen has also to be considered. In a recent publication, Santos et al. [30] have provided calculated values of the Gibbs energy of adsorption of hydrogen on copper single crystals in an aqueous environment. The reported values for Cu (111) and Cu (100) were Δads G ≈ +9.6 kJ mol−1 and +13.5 kJ mol−1, respectively. These Gibbs energies were calculated at the standard potential of the hydrogen evolution reaction, where the positive sign shows that the adsorption is not spontaneous. However, when the overpotential increases (going to more negative potentials), the value of the Gibbs energies decreases and finally becomes negative, making the adsorption of hydrogen possible and more favourable on Cu (111) than on Cu (100). Based on their theoretical work, Santos et al. [30] have found that the first step, the Volmer reaction, is faster on the (111) than on the (100) surfaces. The potential-dependent adsorption of hydrogen and the faster kinetics of the hydrogen evolution reaction at the Cu (111) are fully in agreement with our data reported in Fig. 3 and Table 1 which show that CH4 formation is favoured when the adsorption of hydrogen is promoted. This is consistent with a previous study on Pd–Cu electrodes [31] that demonstrated that the stabilisation of H by Pd markedly increases the methane production.

Fig. 3
figure 3

Amounts of carbon-containing gaseous products detected after the electroreduction of CO2 on copper electrodes. Electrolyte: 0.01 M KH2PO4/0.1 M K2HPO4, electrolysis conditions: E = −1.9 V, t = 3 h, T = 20 °C

Full size image

Electroactivity of Au x Cu100 − x Alloys

The voltammetric measurements of carbon dioxide reduction on the investigated Au x Cu100 − x alloys and Au electrodes are presented in Fig. 4. Because the oxidation of copper–gold alloys results in copper depletion and surface roughening [26], the electrode potential was limited between −0.8 and −1.5 V.

Fig. 4
figure 4

Cyclic voltammograms in CO2-saturated 0.01 M KH2PO4/0.1 M K2HPO4 (solid lines) and N2-saturated 0.1 M KH2PO4/0.02 M K2HPO4 (dashed lines). a Au1Cu99, b Au20Cu80, c Au50Cu50 and d Au. Scan rate 20 mV/s

Full size image

At the Au1Cu99 surface (Fig. 4a), the reduction of CO2 exhibits a clear inhibition of the reaction at potentials more negative than −1.3 V, which is similar to the trend observed at the polycrystalline pure copper surface. Upon increasing the gold content of the Au x Cu100 − x alloys, the inhibition becomes less pronounced (Fig. 4b) then disappears for a 50:50 composition (Fig. 4c). Finally, at the pure gold electrode, the CO2 reduction takes place without inhibition (Fig. 4d). According to these voltammetric results, we can conclude that the presence of gold at the electrode surface facilitates the reduction of CO2. In this respect, it is worth pointing out that the gold content also influences the relative inhibition of the HER by adsorbed CO, as inferred from the comparison of the voltammograms recorded in the presence and absence of CO2.

Examination of Fig. 5 reporting the analytical information obtained after several potentiostatic electrolysis performed at −1.9 V shows that, at the Au x Cu100 − x surfaces, the CO production increases very markedly with the Au content while the fraction of CH4 diminishes. It is known that on pure gold, CO2 electroreduction produces almost exclusively CO in the gaseous phase [32]. Together with the absence of an inhibition feature in the voltammetric curves, this suggests that the adsorption–desorption behaviour of CO on gold and copper substrates might differ considerably from each other. On gold surfaces, the adsorption enthalpies of CO reported in the literature are in the range Δads H ≈ −55 to −59 kJ mol−1 [33], which is not markedly different from the values determined for Cu (poly) or Cu (110). Gottfried et al. [34] have shown that the desorption of carbon monoxide from gold is promoted at high CO coverage, because of the occurrence of dipole–dipole repulsive interactions between adsorbates, but such dependence of the adsorption enthalpy on the coverage is however known in the case of copper surfaces as well [35]. A large difference between gold and copper concerns however the activation energy associated with the CO desorption, which is much lower on gold, E a ≈ +38 kJ mol−1 [34], than on copper, E a ≈ +67 kJ mol−1 [36]. The desorption of CO from gold surfaces is thus kinetically favoured. Due to this easier “CO turnover” (exception made of the Au1Cu99 alloy, for which the results are very similar to those of pure copper), the gold-containing surfaces convert higher quantities of CO2 compared to pure copper that provides the worst results. Amongst all the examined electrodes, the Au50Cu50 alloy appears to be the most efficient substrate for the conversion of CO2 into carbon-containing gaseous products.

Fig. 5
figure 5

Amounts of carbon-containing gaseous products detected after the electroreduction of CO2 on copper, gold and copper–gold surfaces. Electrolyte: 0.01 M KH2PO4/0.1 M K2HPO4, electrolysis conditions: E = −1.9 V, t = 3 h, T = 20 °C

Full size image

To evaluate the actual efficiency of the electrodes, we have to take into account that the production of CO requires only two electrons in contrast to the eight or more electrons needed for the formation of hydrocarbons. The faradaic efficiencies for the reduction of CO2 to gaseous products (Table 2) were calculated by considering the number of electrons required to form each species from CO2 (2 for CO, 8 for CH4 and 12 for C2H4; C2H6 was neglected). Obviously, for all electrodes, only a small part of the total electrolysis charge is useful for the conversion of CO2 into gaseous products, meaning that the most important fraction of the electrolysis charge is consumed by the water reduction side reaction. Moreover, spectroscopic investigations of the electrode|solution interface (results not shown here) revealed, in agreement with previous studies [19], that dissolved CO2 is converted into hydrogen carbonate and carbonate near the electrode due to the alkalinisation of the solution at the electrode when applying negative potentials. The CO2 available for reduction is then considerably diminished, which is unfavourable to the efficiency of the CO2 electrolysis.

Table 2 Faradaic efficiencies of the electrodes for the electrochemical reduction of CO2 in carbon-containing gaseous products
Full size table

Conclusion

At copper electrodes, CO and CH4 are the major products of CO2 electroreduction from aqueous phosphate buffer solutions, their production depending strongly on the potential and the surface morphology. The formation of CH4 occurs at negative potentials where adsorbed H is present at the Cu surface. In a recent publication, Koper et al. [17] introduced a mechanism for the electrochemical reduction of CO2 on copper based on two pathways, one for C1 compounds and another for C2 compounds, with CO(ads) as common intermediate. From our results we conclude that the pathway to CH4 takes place indeed via reactions between adsorbed CO and H. This was also previously suggested by Dubé and Brisard [37], who proposed a mechanism in which adsorbed molecules react at the electrode surface to form hydrogenated species similarly to an electrocatalytic hydrogenation process. To react, the adsorbed species have to be located on adjacent sites. The formation of CH4 is thus expected to be promoted on high-density flat surfaces, like Cu (111). This interpretation is in line with the analysis of the reaction products after long-term electrolysis experiments. It should however be emphasized that at the electrolysis potential considered here, H(ads) is also consumed by the competitive hydrogen evolution reaction, limiting the amount of adsorbed H available to react with adsorbed CO, and making H2 the main “side product” of the reduction. C2H4 is the major C2 compound produced in our conditions, while only traces of C2H6 and C2H5OH were detected. However, in contrast to the Koper group [17], we cannot exclude the other routes proposed for C2 compounds by Hori [38], i.e. the combination of two adsorbed CH2 species and the combination of adsorbed CO and CH2, even if they occur in a small extent, as we detected traces of C2H6 and C2H5OH. Performing long-time electrolysis and accumulating the products before chromatographic analysis have allowed us probably to detect the presence of C2H6 and C2H5OH in very small quantities that might be undetectable during real-time experiments as were performed by Koper et al. Moreover, the potentiostatic conditions used are favourable to the CH4 formation, and thus, the CH2(ads) coverage is most likely sufficient to permit the combination reactions to occasionally occur.

Efficient CO-producing electrodes could be obtained by alloying copper with gold. The dilution of the copper sites at the surface readily prevents the formation of gaseous hydrogenated species and activates the production of CO. Among all the investigated compositions, the alloy composed of equal quantities of copper and gold gives the best results. As mentioned before, the CO desorption is kinetically favoured on Au compared to Cu, and the presence of adsorbed CO promotes its desorption. We thus suggest that, at the Au50Cu50 surface, the presence of CO(ads) blocked on Cu sites, which represent on average one atom out of two, reinforces the desorption of CO from the Au sites, accelerating the turnover at these sites.